Files
claudecodeui/server/modules/providers/README.md
Haile 4cee5e7286 Fix/resolve different bugs (#964)
* fix: don't overlap thinking banner over conversation

* feat: support message queue and add attention indicator

* fix(shell): use c to copy for claude

* fix: strip timestamp tags from cursor

* fix: select project when loading session from direct URL

* feat: support attached images for all providers

* feat(opencode): support permission options

* fix: resolve source control and chat UX bugs

- make staging real in the git panel: new /stage and /unstage endpoints,
  /status now reports the actual index via porcelain -z (handles renames,
  conflicts, and paths with spaces), and Stage/Unstage All run real git
  commands instead of toggling client-side state
- add branch search to the header dropdown and Branches tab
- persist the chosen permission mode per provider so a brand-new chat
  keeps it once the session id arrives, instead of reverting to default
- replace cmdk's fuzzy model search with strict token matching so
  "chatgpt" no longer surfaces unrelated models
- unit tests for the git status parser

* feat(git): commit graph in history view and cross-spawn everywhere

- render a VSCode-style commit graph in the source control history:
  lane-assignment algorithm, SVG strip with colored rails, merge and
  branch curves, commit dots, and branch/tag badges per commit
- /commits returns parents and ref decorations across all branches
  (--branches --remotes --tags --topo-order) using unit-separator
  fields so pipes in commit subjects can't break parsing
- collect stats via a single --shortstat pass instead of one
  `git show --stat` call per commit; raise the history limit to 50
- replace child_process spawn with cross-spawn in all runtimes,
  routes, and services: it resolves .cmd shims and PATHEXT on
  Windows (fixing taskmaster's npx invocations) and delegates to
  native spawn elsewhere, removing the per-file win32 ternaries
- unit tests for the log parser and lane assignment

* fix: remove gemini support since google discontinued it

* fix: address code review findings

- validate chat image attachments server-side: only files inside the
  ~/.cloudcli/assets upload store may reach provider file reads
- harden asset serving with nosniff and attachment disposition for
  SVGs to prevent stored XSS
- show the timestamp on image-only user messages
- serialize git stage/unstage calls and defer the status re-sync so
  rapid toggles can't interleave or flicker
- use a literal hex fallback for commit ref badges (alpha suffix on a
  var() string produced invalid CSS)
- stop binding a URL session to a guening
  project instead
- add the missing attentionRequiredIndicator key to all sidebar locales
- clean up dangling conjunctions left the
  de/ru/tr/ja/zh-CN/zh-TW READMEs

* fix: address code scanning findings

- enforce a second image trust boundary in the provider builders:
  buildClaudeUserContent/buildCodexInputItems only accept files inside
  the upload store or the run's working directory (CodeQL path
  injection)
- drop an always-true selectedProject check in handleSubmit
- remove the unused resume option from spawnCursor

* fix: use CodeQL-recognized containment check for image reads

Replace the path.relative guard in isPathInsideDirectory with the
resolve + startsWith(root + sep) idiom so the path-injection barrier
is visible to code scanning; behavior is unchanged.

* fix(security): canonicalize Claude image attachment reads

Resolve image attachment paths with realpath before reading so symlinks
inside allowed roots cannot escape to arbitrary files. Preserve support for
symlinked project/upload roots and add focused coverage for both cases.

* fix: show agent subtask

* fix(sessions): title app-created sessions from the first user message

App-created sessions (started by sending a message from cloudcli) were
titled with placeholder names — "Untitled Codex Session" from the disk
indexer and, briefly, "New session" from the empty canonical upsert —
before a later sync finally settled on the right text, causing the
sidebar title to flicker.

- Codex/OpenCode synchronizers now title app-created sessions (distinct
  app id mapped to a provider id) from the first user message, while
  sessions found purely by indexing keep their existing setup. Claude
  keeps its AI-generated titles; Cursor already used the first message.
- Decode OpenCode's JSON-string-literal prompts so titles no longer
  surface wrapped in quotes; hoist unwrapJsonStringLiteral into
  shared/utils since it's now used by both the reader and synchronizer.
- Guard the sidebar upsert merge so an empty summary can never blank out
  a title that is already set.
- Add codex/opencode synchronizer tests for the app-created vs indexed
  naming paths.

* fix(chat): make message queuing reliable across sessions and turn boundaries

Queued messages had four related defects:

- A queued message flashed and then vanished at flush time: concurrent
  transcript refreshes (the `complete` handler racing the watcher-triggered
  update) could resolve out of order, letting a stale response overwrite
  newer server messages after the optimistic row was pruned. Session slots
  now carry a monotonic fetch ticket and discard stale fetch/refresh/
  fetchMore responses.

- Switching sessions flushed the previous session's queued draft into the
  newly viewed one (sending it with the wrong provider's settings, e.g. a
  Claude model into a Codex session). The composer flush is now scoped to
  its session, and a draft restored into an idle session sends after a
  short grace period so a live-run ack can cancel it.

- The thinking banner never appeared for a queued turn: the chat handler's
  session-keyed completeRun safety net could fire after a queued message
  had already started the session's next run, emitting a spurious
  `complete` that killed it. The safety net is now scoped to its own run
  via completeRunIfCurrent (with a regression test).

- Queued messages only sent while their session was being viewed. Drafts
  now persist their send options (model, effort, permissions) at queue
  time, and a new app-level useQueuedMessageAutoSend hook dispatches a
  non-viewed session's queued message as soon as its run completes, using
  the storage key as the claim ticket to prevent double sends.
2026-07-08 10:32:32 +02:00

14 KiB

Providers Module Guide

This file documents the current provider contract in server/modules/providers. Keep it current whenever provider wiring, skill discovery, or session sync behavior changes. The goal is that a human or AI agent can add a new provider without guessing which files need to move.

Current Provider Shape

Every provider wrapper exposes five facets:

  • auth
  • mcp
  • skills
  • sessions
  • sessionSynchronizer

These correspond to the shared interfaces in server/shared/interfaces.ts:

  • IProviderAuth
  • IProviderMcp
  • IProviderSkills
  • IProviderSessions
  • IProviderSessionSynchronizer

The services that consume them are:

  • providerAuthService
  • providerMcpService
  • providerSkillsService
  • sessionsService
  • sessionSynchronizerService

Current provider ids in this repo are:

  • claude
  • codex
  • cursor
  • opencode

Those ids are mirrored in backend unions and frontend provider constants. If adding a new provider, update every place that hardcodes this list.

Current File Layout

Each provider lives under its own folder in server/modules/providers/list/:

server/modules/providers/list/<provider>/
  <provider>.provider.ts
  <provider>-auth.provider.ts
  <provider>-mcp.provider.ts
  <provider>-skills.provider.ts
  <provider>-sessions.provider.ts
  <provider>-session-synchronizer.provider.ts

The existing provider folders are claude, codex, cursor, and opencode.

What Each Facet Does

Facet Responsibility Base / Service
auth Report install/auth state for the provider runtime IProviderAuth -> providerAuthService
mcp Read, list, write, and remove provider-native MCP config McpProvider -> providerMcpService
skills Discover provider-native skill markdown files SkillsProvider -> providerSkillsService
sessions Normalize live events and fetch session history IProviderSessions -> sessionsService
sessionSynchronizer Scan transcript artifacts and upsert session metadata IProviderSessionSynchronizer -> sessionSynchronizerService

sessions and sessionSynchronizer are separate concerns:

  • sessions handles runtime event normalization and history fetches.
  • sessionSynchronizer handles file-backed session indexing into sessionsDb.

How To Add A Provider

  1. Add the provider id everywhere it is part of the contract.
  • Update server/shared/types.ts LLMProvider.
  • Update src/types/app.ts LLMProvider if the frontend should know about it.
  • Update server/modules/providers/provider.routes.ts.
  • Update server/routes/agent.js if the provider is launchable from the agent runtime.
  • Update server/index.js if the provider needs runtime boot or shutdown wiring.
  • Update the PROVIDER_ORDER list in public/api-docs.html if the provider should appear in the public API docs.
  • Update src/components/chat/hooks/useChatProviderState.ts and src/components/chat/view/subcomponents/ProviderSelectionEmptyState.tsx if the provider should be selectable in chat.
  • Update src/components/provider-auth/view/ProviderLoginModal.tsx if the provider has a login/setup flow.
  1. Create the wrapper class.
  • Add server/modules/providers/list/<provider>/<provider>.provider.ts.
  • Extend AbstractProvider.
  • Expose readonly auth, mcp, skills, sessions, and sessionSynchronizer.
  • Call super('<provider>').
  1. Implement auth.
  • Return a full ProviderAuthStatus.
  • Treat normal not installed / not authenticated states as data, not exceptions.
  • Keep provider-specific credential discovery inside the auth provider.
  • If the provider has no auth step, return a stable unauthenticated or not-installed status instead of omitting the facet.
  1. Implement MCP.
  • Extend McpProvider.
  • Pass the supported scopes and transports to super(...).
  • Implement the four required methods:
    • readScopedServers(...)
    • writeScopedServers(...)
    • buildServerConfig(...)
    • normalizeServerConfig(...)
  • Use the shared validation and normalization behavior from McpProvider.
  • Keep the provider-specific config format local to the provider implementation.

Current MCP formats in this repo are:

Provider User / Project Storage Supported Scopes Supported Transports
Claude .mcp.json in user / local / project locations user, local, project stdio, http, sse
Codex .codex/config.toml user, project stdio, http
Cursor .cursor/mcp.json user, project stdio, http
OpenCode ~/.config/opencode/opencode.json or <workspace>/opencode.json (.jsonc is read when present) user, project stdio, http
  1. Implement skills.
  • Extend SkillsProvider.
  • Implement getSkillSources(workspacePath).
  • Return the actual discovery roots for the provider.
  • Skills are discovered from SKILL.md files.
  • readProviderSkillMarkdownDefinition(...) reads front matter name and description.
  • If name is missing, the parent directory name is used as a fallback.
  • Use recursive: true only when the provider stores skills in nested trees.
  • Keep the emitted command string aligned with the provider's real skill syntax.

Current skill discovery roots are:

Provider User Roots Project / Repo Roots Prefix Notes
Claude ~/.claude/skills <workspace>/.claude/skills / Also discovers Claude plugin skills from enabled plugin installs. Command skills live under commands/; markdown skills live under skills/ and are scanned recursively.
Codex ~/.agents/skills, ~/.codex/skills/.system, /etc/codex/skills <workspace>/.agents/skills, path.dirname(workspacePath)/.agents/skills, topmost git root .agents/skills $ Overlapping roots are deduplicated before scanning.
Cursor ~/.cursor/skills <workspace>/.cursor/skills, <workspace>/.agents/skills / Uses slash-style commands.
OpenCode ~/.config/opencode/skills, ~/.claude/skills, ~/.agents/skills Cwd-to-topmost-git-root .opencode/skills, .claude/skills, and .agents/skills / Reuses OpenCode, Claude, and Agents skill locations. Overlapping roots are deduplicated before scanning.

Command forms currently used by the providers are:

  • Claude user/project skills: /skill-name
  • Claude plugin skills: /plugin-name:skill-name
  • Codex skills: $skill-name
  • Cursor skills: /skill-name
  • OpenCode skills: /skill-name
  1. Implement sessions.
  • Implement normalizeMessage(raw, sessionId) and fetchHistory(sessionId, options).
  • Use createNormalizedMessage(...) and generateMessageId(...) for emitted messages.
  • Keep normalized message ids unique. If one raw event produces multiple text parts, append a discriminator so ids do not collide.
  • Keep pagination consistent:
    • limit: null means unbounded/full history.
    • limit: 0 means an empty page.
    • always return total, hasMore, offset, and limit when paginating.
  • Sanitize any filesystem-derived ids before using them in file or database paths.
  • Do not assume a provider's history format matches another provider's format.
  1. Implement session synchronization.
  • Implement synchronize(since?: Date) to scan provider artifacts and upsert sessions into sessionsDb.
  • Implement synchronizeFile(filePath) for single-file watcher updates.
  • Use the existing helpers when they fit:
    • buildLookupMap(...)
    • extractFirstValidJsonlData(...)
    • findFilesRecursivelyCreatedAfter(...)
    • normalizeSessionName(...)
    • readFileTimestamps(...)
  • Make the sync resilient to partial, malformed, or missing provider files.
  • The orchestration service runs all provider synchronizers and only advances scan_state.last_scanned_at when every provider succeeds.

Current session sync roots are:

Provider Scan Roots Metadata Helpers / Notes
Claude ~/.claude/projects/**/*.jsonl Uses ~/.claude/history.jsonl for name lookup and the trailing ai-title, last-prompt, or custom-title entries for title recovery.
Codex ~/.codex/sessions/**/*.jsonl Uses ~/.codex/session_index.jsonl for title lookup and the last task_complete message for a fallback title.
Cursor ~/.cursor/projects/**/*.jsonl Uses sibling worker.log to recover workspacePath, then derives the session title from the first user prompt.
OpenCode ~/.local/share/opencode/opencode.db Reads active sessions/messages/parts from OpenCode's shared SQLite database and stores jsonl_path as null so deleting one app session cannot remove the shared DB.
  1. Register the provider.
  • Add the new provider class to server/modules/providers/provider.registry.ts.
  • Update server/modules/providers/provider.routes.ts provider parsing.
  • If the provider introduces a new service or lifecycle hook, export it from the module entrypoint that consumes providers.
  1. Wire runtime and UI surfaces outside the providers module when needed.

If the provider can run live chat sessions, update the runtime entrypoints too:

  • server/routes/agent.js
  • server/index.js

If the provider is visible in the UI, update:

  • provider model fallback files under server/modules/providers/list/<provider>/
  • src/components/chat/hooks/useChatProviderState.ts
  • src/components/chat/view/subcomponents/ProviderSelectionEmptyState.tsx
  • src/components/provider-auth/view/ProviderLoginModal.tsx
  • src/components/mcp/constants.ts

Minimal Wrapper Template

import { AbstractProvider } from '@/modules/providers/shared/base/abstract.provider.js';
import { <Provider>ProviderAuth } from './<provider>-auth.provider.js';
import { <Provider>McpProvider } from './<provider>-mcp.provider.js';
import { <Provider>SkillsProvider } from './<provider>-skills.provider.js';
import { <Provider>SessionsProvider } from './<provider>-sessions.provider.js';
import { <Provider>SessionSynchronizer } from './<provider>-session-synchronizer.provider.js';
import type {
  IProviderAuth,
  IProviderMcp,
  IProviderSessionSynchronizer,
  IProviderSessions,
  IProviderSkills,
} from '@/shared/interfaces.js';

export class <Provider>Provider extends AbstractProvider {
  readonly auth: IProviderAuth = new <Provider>ProviderAuth();
  readonly mcp: IProviderMcp = new <Provider>McpProvider();
  readonly skills: IProviderSkills = new <Provider>SkillsProvider();
  readonly sessions: IProviderSessions = new <Provider>SessionsProvider();
  readonly sessionSynchronizer: IProviderSessionSynchronizer =
    new <Provider>SessionSynchronizer();

  constructor() {
    super('<provider>');
  }
}

Minimal Skills Template

import path from 'node:path';

import { SkillsProvider } from '@/modules/providers/shared/skills/skills.provider.js';
import type { ProviderSkillSource } from '@/shared/types.js';

export class <Provider>SkillsProvider extends SkillsProvider {
  constructor() {
    super('<provider>');
  }

  protected async getSkillSources(workspacePath: string): Promise<ProviderSkillSource[]> {
    return [
      {
        scope: 'project',
        rootDir: path.join(workspacePath, '.<provider>', 'skills'),
        commandPrefix: '/',
      },
    ];
  }
}

Minimal Session Sync Template

import type { IProviderSessionSynchronizer } from '@/shared/interfaces.js';

export class <Provider>SessionSynchronizer implements IProviderSessionSynchronizer {
  async synchronize(since?: Date): Promise<number> {
    return 0;
  }

  async synchronizeFile(filePath: string): Promise<string | null> {
    return null;
  }
}

AI Prompt Template

Use this prompt when asking an AI agent to add a provider:

Add a new provider "<provider>" using the current provider module architecture.

Requirements:
1) Create:
   - server/modules/providers/list/<provider>/<provider>.provider.ts
   - server/modules/providers/list/<provider>/<provider>-auth.provider.ts
   - server/modules/providers/list/<provider>/<provider>-mcp.provider.ts
   - server/modules/providers/list/<provider>/<provider>-skills.provider.ts
   - server/modules/providers/list/<provider>/<provider>-sessions.provider.ts
   - server/modules/providers/list/<provider>/<provider>-session-synchronizer.provider.ts
2) Register in:
   - server/modules/providers/provider.registry.ts
   - server/modules/providers/provider.routes.ts
   - server/shared/types.ts LLMProvider
   - src/types/app.ts LLMProvider
3) Mirror the nearest existing provider implementation for file naming, style,
   and error handling.
4) Implement skills support with SkillsProvider and the current skill roots.
5) Implement session synchronization if the provider stores transcript files.
6) Ensure sessions use unique ids, safe path handling, and correct pagination.
7) Keep `sessions` and `sessionSynchronizer` separate.
8) Run:
   - npx eslint <touched files>
   - npx tsc --noEmit -p server/tsconfig.json

Validation

After adding or changing a provider, run the relevant checks:

npx eslint server/modules/providers/**/*.ts server/shared/types.ts server/shared/interfaces.ts
npx tsc --noEmit -p server/tsconfig.json

Useful tests in this repo:

  • server/modules/providers/tests/mcp.test.ts
  • server/modules/providers/tests/skills.test.ts
  • server/modules/providers/tests/opencode-sessions.test.ts

If you touch sessions or session synchronization, add or update focused tests alongside the implementation.

Common Mistakes

  • Adding provider files but forgetting provider.registry.ts or provider.routes.ts.
  • Updating backend provider ids but not src/types/app.ts or the frontend provider constants.
  • Omitting skills or sessionSynchronizer from the wrapper.
  • Returning duplicate normalized message ids for split content.
  • Treating limit === 0 as unbounded history.
  • Building file paths from raw session ids without validation.
  • Hardcoding a skill root without checking the provider's actual discovery rules.
  • Forgetting that Claude plugin skills are discovered differently from normal user/project skill folders.
  • Assuming one provider's MCP config file format works for the others.